1. Introduction

This is an R Markdown document that consolidates my exploratory data analysis (EDA) and regression analysis to understand the relationship between citizens’ participation and fiscal outcomes.

This study entailed the construction of a data set which combines local budget and fiscal performance data with available measures of local participation and other crucial covariates. The data set constructed leverages on the work already done by the World Bank (2021) and by C. Magno et al. (2022), mainly by enhancing the indicators on citizen participation.

This study adopts a panel regression model specified as follows:

y_hjt = D_jt’δ + X_ijt’β + u_jt

where y_hjt stands for budget performance measured by the hth metric of the jth Local Government Unit (LGU) in year t; D_jt is the vector of measures of participatory governance; X_ijt is the vector of covariates on the governance and socioeconomic characteristics of LGUs, grouped in i dimensions; and u_jt is the error term, decomposed further as the sum of α_j (LGU fixed effects) and ε_jt (idiosyncratic error).

The dependent variables y_hjt are metrics that illustrate the achievement of good public financial management (PFM): fiscal discipline, allocative efficiency, and operational efficiency. The explanatory variables D_jt are indicators that are available from the Seal of Good Local Governance (SGLG) on the participation of Civil Society Organizations (CSOs) in Local Development Councils (LDCs). The other covariates X_ijt are adapted from C. Magno et al. (2022) and include size of aggregate and functional budgets, LGU performance as measured by the SGLG, and socioeconomic development.

Based on the study’s empirical strategy, additional processing needed to be done in order to prepare the data for the regression analyses. First, only observations from 2017-2018, when all the CSO participation variables are present, will be included in the regression analyses. Second, extreme outliers and other defects in the dependent variables and one of the continuous explanatory variables (proportion of CSOs in the LDC) will be normalized through Winsorization at 98% (i.e., values below 1st percentile and above 99th percentile will be replaced with “NA”). Note: tinker the code below to change the percentile cutoffs.

2. Exploration of the Variables

2.1 Outcome Variables (y_hjt)

The first outcome of good PFM is fiscal discipline, which pertains to the ability to keep government expenditure consistent with revenues. Typically, the fiscal deficit–the gap between revenues and disbursements, and the amount that needs to be borrowed to fill such gap–is used as the main metric of fiscal discipline. This is not applicable in the context of Philippine LGUs, which are enjoined by the LDC to maintain a balanced budget.

For the purpose of this study, metrics on the actual means of the LGUs will be used to measure performacne in this outcome area: (a) year on year growth of locally generated revenue and (b) dependence on fiscal transfers particularly the National Tax Allocation (NTA). On the former, annual growth of local revenues averaged 18% from 2017 to 2018. Although municipalities experienced the highest growth–an average of 19%%–cities continue to generate significantly higher revenues per capita compared to municipalities and provinces.

Summary of y1 = Annual Local Revenue Growth
Mean St Dev
City 0.1327 0.1638
Municipality 0.1912 0.3810
Province 0.1458 0.3758
All LGUs 0.1841 0.3676

Meanwhile, NTA dependence–which pertains to how LGUs rely on transfers from the national government rather than revenues they generate on their own–continued to be high at 80%. Municipalities and provinces are more NTA dependent at 82% and 80%, respectively, compared to 62% among cities which are more capable of generating revenues from real property, business, and other taxes. Cities and Municipalities are relatively on par when it comes the NTA they receive per capita, implying that Cities, which have higher per capita revenues, have more resources at their disposal.

Summary of y2 = NTA Dependence
Mean St Dev
City 0.6153 0.2424
Municipality 0.8177 0.1838
Province 0.7987 0.1253
All LGUs 0.7994 0.1955

The second PFM outcome–allocative efficiency–pertains to the ability of LGUs to allocate more resources to development priorities. While LGUs need to spend according to their unique local priorities, one could infer that it is better for LGUs to increase the share of productive expenditures, as opposed to overhead and other expenditures not for socioeconomic development.

There are two ways that one could measure productive expenditures. The first is the combined shares of budgets for maintenance and other operating expenditures (MOOE) and for capital outlays (CO) in the total budget. The remainder are budgets for personnel services (PS), debt servicing, and other unclassified items. Based on this definition, 60% of LGUs’ budgets, on average, are allocated for MOOE and CO, with provinces allocating largest shares of about 68%.

Summary of y3 = Share of Productive Expenditures (MOOE + CO)
Mean St Dev
City 0.6786 0.1211
Municipality 0.5917 0.1334
Province 0.6844 0.1043
All LGUs 0.6042 0.1346

The second way of calculating for productive expenditures is by sector and fund, specifically by excluding general public services (GPS) and other sectors from the total. This results in a lower share of productive expenditures at an average of 41%. Still, provinces continue to allocate larger shares to productive expenditures compared to municipalities.

Summary of y4 = Share of Productive Sectors in the Budget
Mean St Dev
City 0.4799 0.1467
Municipality 0.3986 0.1404
Province 0.5620 0.1305
All LGUs 0.4143 0.1464

Finally, operational efficiency pertains to the ability of LGUs to utilize resources effectively to deliver services. For this study, we use two indicators of operational efficiency: (a) per capita expenditures, which is a rough proxy for LGUs’ overall capability to deliver services, and (b) the budget utilization rate (BUR), which stands for the rate of actual expenditures against the approved budgets of LGUs.

LGUs per capita expenditures average PHP 3,274 annually. As expected, cities spend more per capita at PHP 4,742. Meanwhile, provinces spend the least per person at PHP 2,129. Growth in per capita expenditures also affects per capita productive spending, which averages PHP 1,301 annually.

Summary of y5 = Per Capita Expenditures
Mean St Dev
City 4742.352 2591.596
Municipality 3193.673 2226.398
Province 2129.520 1184.122
All LGUs 3274.375 2277.186
Summary of y5 = Per Capita Productive Expenditures
Mean St Dev
City 2046.686 1796.567
Municipality 1233.327 1800.571
Province 1217.127 1091.459
All LGUs 1301.638 1786.623

The budget utilization rate (BUR) has been the subject of the study by C. Magno et al. (2022). On average, LGUs utilize about 78% of their budgets. Cities tend to have lower BUR at 69%, while municipalities’ are higher at 79%. The overall BUR of LGUs have been decreasing. Finally, the LGUs’ utilization of their budgets for productive sectors is lower at 72%.

Summary of y6 = Budget Utilization Rate
Mean St Dev
City 0.6942 0.1680
Municipality 0.7882 0.1479
Province 0.7196 0.1443
All LGUs 0.7764 0.1525
Summary of y6 = Budget Utilization Rate
Mean St Dev
City 0.6817 0.1828
Municipality 0.7267 0.3035
Province 0.6772 0.1973
All LGUs 0.7202 0.2906

2.2 Explanatory Variables (D_jt)

Beginning 2017, measures on the functionality of LDCs have been added to the SGLG. These measures include whether the Council was organized and if CSO members of the council meet or exceed the 25% requirement of the LGC. C. Magno et al. (2022) found that the share of CSO members in the LDC did not have a statistically significant relationship with LGUs’ BUR. This thesis replicates the said analysis by relating CSO participation with other fiscal outcomes and making use of other variables available from the SGLG on CSO participation in LDCs:

  1. Whether or not the LGU met the minimum 25% of CSO members in the LDC

  2. Actual proportion of CSO members in the LDC

  3. Whether or not the LDC secretariat tapped the CSO (non-government organizations, research organizations and/or academic institutions) for technical support.

  4. Whether or not the CSO members participated in the LDC deliberations

  5. Whether or not the CSO members submitted a CSO plan to follow through on LDC deliberations

The indicators outlined above could be considered as measuring varying depths of CSO participation: from mere presence (where nearly all LGUs claim to have the minimum CSO representatives in the LDC) to a more meaningful engagement (where only in a minority of LGUs have CSOs submitted an action plan). Though the indicators do not reveal the substance or topics of the discussions, the quality of the action plans, or even the power dynamics between the local officials and the CSOs, these indicators are useful for now in illustrating the funnel of citizen participation.

Summary of SGLG Indicators on CSO Participation in LDCs
LDCs with Minimum CSO Members Mean % CSO Representation LDC Secretariats with CSO Support LDCs with CSOs Attending Meetings LDCs with CSOs Submitting Action Plans
City 0.8493 0.2811 0.4966 0.8390 0.1884
Municipality 0.7823 0.3038 0.4685 0.7634 0.1652
Province 0.9074 0.3133 0.5679 0.8642 0.1605
All LGUs 0.7939 0.3004 0.4756 0.7746 0.1670

In addition to participation in the LDC, the SGLG also has indicators on the participation of CSOs (a minimum of four representatives, plus minimum of one other member from the private sector) in the Local Risk Reduction and Management Councils (LDRRMCs). Under these shallow metrics, 84 percent of LGUs have met the minimum requirement.

Summary of SGLG Indicators on CSO Participation in LDRRMC
LDRRMCs with Minimum CSO Members LDRRMCs with Minimum Private Sector Member
City 0.8801 0.8870
Municipality 0.8325 0.8208
Province 0.9259 0.9259
All LGUs 0.8410 0.8314

It can be argued that meaningful participation cannot be achieved without proper mechanisms to access the information they need and to hold LGUs accountable.

On transparency, the SGLG tracks LGUs’ compliance with the Full Disclosure Policy (FDP) where LGUs are required to post financial and other reports physically in conspicuous places in the provincial, city, or municipal hall; and to submit the same in the FDP Portal. Compliance is high at about 90%.

Summary of SGLG Indicators on Full Disclosure
LGUs that Conspiculously Post their Financial Disclosures LGUs that Submit Financial Reports to the FDP Portal
City 0.9144 0.9075
Municipality 0.8942 0.8581
Province 0.9630 0.9444
All LGUs 0.8992 0.8664

Meanwhile, accountability is understood in many ways–from having strong oversight from the legislature and an independent audit office, to the ultimate accountability mechanism of the elections–but for the purpose of this study we use the audit rating provided by the Commission on Audit (COA) on the financial accounts of LGUs. As could be gleaned below, an overwhelming 88% of LGUs have qualified audit findings.

Summary of LGU Audit Ratings
Unqualified Qualified Disclaimer Adverse No Opinion or Report
City 0.0377 0.9041 0.0103 0.0411 0.0068
Municipality 0.0738 0.8804 0.0199 0.0108 0.0152
Province 0.1296 0.8519 0.0062 0.0000 0.0123
All LGUs 0.0734 0.8810 0.0184 0.0129 0.0143

3. Exploration of Relationships

The scatterplots shown below seem to imply no strong relationship between CSO participation measured as proportion of CSO members in the LDC and good budget performance. These are with the expception of the share of productive sectors and funds in the budget (slightly negative) and spending per capita (positive).

## `geom_smooth()` using formula = 'y ~ x'

## `geom_smooth()` using formula = 'y ~ x'

## `geom_smooth()` using formula = 'y ~ x'

## `geom_smooth()` using formula = 'y ~ x'

## `geom_smooth()` using formula = 'y ~ x'

## `geom_smooth()` using formula = 'y ~ x'